Introduction

“Our mission is to build AI responsibly to benefit humanity”1About Google DeepMind, Google DeepMind, (<)a href='https://deepmind.google/about/'(>)https://deepmind.google/about/(<)/a(>) (last visited Oct. 25, 2025). – Google DeepMind

“Our primary fiduciary duty is to humanity”2OpenAI Charter, (<)a href='https://openai.com/charter/'(>)https://openai.com/charter/(<)/a(>) (last visited Oct. 25, 2025). – OpenAI 

“The specific public benefit that the Corporation will promote is to responsibly develop and maintain advanced AI for the long-term benefit of humanity”3Anthropic Certificates of Incorporation, (<)a href='https://drive.google.com/drive/folders/1ImqXYv9_H2FTNAujZfu3EPtYFD4xIlHJ'(>)https://drive.google.com/drive/folders/1ImqXYv9_H2FTNAujZfu3EPtYFD4xIlHJ(<)/a(>), [(<)a href='https://perma.cc/9TGX-PRGZ'(>)https://perma.cc/9TGX-PRGZ(<)/a(>)] (last visited Oct. 25, 2025). – Anthropic

Leading AI companies all purport to “benefit humanity.” Though each company frames its mission slightly differently,4About Google DeepMind, (<)em(>)supra(<)/em(>) note 1 (“OUR MISSION Build AI responsibly to benefit humanity”); (<)em(>)see also(<)/em(>) Sundar Pichai, (<)em(>)AI at Google: Our Principles(<)/em(>), Google (June 7, 2018), (<)a href='https://www.blog.google/technology/ai/ai-principles/'(>)https://www.blog.google/technology/ai/ai-principles/(<)/a(>), [(<)a href='https://perma.cc/3ALF-WTEC'(>)https://perma.cc/3ALF-WTEC(<)/a(>)]; for the full description of OpenAI’s mission, see OpenAI, (<)em(>)Internal Revenue Service Form 1023: Application for Recognition of Exemption Under Section 501(c)(3) of the Internal Revenue Code(<)/em(>) 31 (2016), (<)a href='https://www.documentcloud.org/documents/25197523-openai-application-for-tax-exempt-status/'(>)https://www.documentcloud.org/documents/25197523-openai-application-for-tax-exempt-status/(<)/a(>) (stating that OpenAI’s “goal is to advance digital intelligence in the way that is most likely to benefit humanity as a whole”)(<)em(>); see also OpenAI Charter, supra(<)/em(>) note 2 (describing OpenAI’s goal using identical language); Anthropic Certificates of Incorporation, (<)em(>)supra(<)/em(>) note 3 (describing Anthropic’s proposed “specific public benefit” as being “to responsibly develop and maintain advanced AI for the long-term benefit of humanity”).(<)br(>) at a high level they signal remarkably similar public-minded aspirations. This report addresses whether the distinct business structures of major AI companies have any bearing on their concrete actions. The legal framework under which these U.S.-based AI businesses operate varies among them and over time, offering an opportunity to compare how different business structures align mission with action. Google/Alphabet is a traditional for-profit corporation; OpenAI morphed from a non-profit to a hybrid ‘capped-profit’ and recently to a public benefit corporation seeking to combine profit and social objectives, whereas Anthropic combines a public benefit corporation with a long-term benefit trust. Yet there is one striking commonality across leading AI companies: they all proceed from the thesis that the broad societal impact of AI is too important to be left to purely corporate incentives. At the end of the day, however, they all ultimately move to the beat of the market, resolving any tensions in favor of financial interest.

AI companies diverge as to how they structure their activities to achieve their mission, but ultimately converge towards for-profit goals. AI being a capital-intensive industry, companies need to fundraise to sustain the ever-rising costs of training and running AI systems.5(<)em(>)Deals – Anthropic – Company Profile(<)/em(>), (<)a href='https://my.pitchbook.com/profile/466959-97/company/deals#deal-287700-58T'(>)https://my.pitchbook.com/profile/466959-97/company/deals#deal-287700-58T(<)/a(>) (last visited Sep. 11, 2025) (showing that Anthropic raised over $30 billion to date, with a $13 billion funding round in September 2025). Notwithstanding the shaky fundamentals, these companies still achieve astronomical valuations and investor interest.6Despite revenue growth, profit often remains elusive; (<)em(>)PitchBook Profile – OpenAI(<)/em(>), (<)a href='https://content.pitchbook.com/share/quick-link/profile/9374fa6e-c6a6-4a7f-8402-fbc780b08e48?hash=8fac7125145ad4bf2fc8a3a6fc1239b49457cc22a3ed196a5020b3b081e26fa9'(>)https://content.pitchbook.com/share/quick-link/profile/9374fa6e-c6a6-4a7f-8402-fbc780b08e48?hash=8fac7125145ad4bf2fc8a3a6fc1239b49457cc22a3ed196a5020b3b081e26fa9(<)/a(>) (last visited June 23, 2025) (showing that OpenAI lost $5 billion in 2025, despite drawing a revenue upwards of $3.7 billion and a net profit margin of -134% in 2024); Cory Weinberg, (<)em(>)OpenAI Projections Imply Losses Tripling to $14 Billion in 2026(<)/em(>), Information (Oct. 9, 2024), (<)a href='https://www.theinformation.com/articles/openai-projections-imply-losses-tripling-to-14-billion-in-2026'(>)https://www.theinformation.com/articles/openai-projections-imply-losses-tripling-to-14-billion-in-2026(<)/a(>) (last visited Oct. 27, 2025); (<)em(>)Will OpenAI Ever Make Real Money?(<)/em(>), Economist (May 15, 2025), (<)a href='https://www.economist.com/business/2025/05/15/will-openai-ever-make-real-money'(>)https://www.economist.com/business/2025/05/15/will-openai-ever-make-real-money(<)/a(>) (last visited June 2, 2025); Mike Isaac & Erin Griffith, (<)em(>)OpenAI Is Growing Fast and Burning Through Piles of Money(<)/em(>), N.Y. Times (Sep. 27, 2024), (<)a href='https://www.nytimes.com/2024/09/27/technology/openai-chatgpt-investors-funding.html'(>)https://www.nytimes.com/2024/09/27/technology/openai-chatgpt-investors-funding.html(<)/a(>); Berber Jin, (<)em(>)Mind-Blowing Growth Is About to Propel Anthropic Into Its First Profitable Quarter(<)/em(>), Wall St. J. (May 20, 2026), (<)a href='https://www.wsj.com/tech/ai/mind-blowing-growth-is-about-to-propel-anthropic-into-its-first-profitable-quarter-7edbf2f4'(>)https://www.wsj.com/tech/ai/mind-blowing-growth-is-about-to-propel-anthropic-into-its-first-profitable-quarter-7edbf2f4(<)/a(>) (describing unaudited accounting projects in Anthropic’s first profitable quarter, but noting that profits are not expected to last given the company’s infrastructure spending); Anthropic, PBC, Amendment No. 1 to Registration Statement (Form S-1/A) 23 (Nov. 2, 2026), (<)a href='https://www.sec.gov/Archives/edgar/data/2130357/000162828026068676/anthropicpbc-sx1ax1.htm'(>)https://www.sec.gov/Archives/edgar/data/2130357/000162828026068676/anthropicpbc-sx1ax1.htm(<)/a(>) (disclosing a $1.26 billion operating loss on $33.2 billion revenue in the first nine months of 2026). In light of this financial pressure, it’s unclear whether any of them have the right incentives to prioritize aspirational goals to “benefit humanity” when that would adversely impact their bottom line. The 244-billion-dollar question—the estimated value of the current AI market7(<)em(>)Artificial Intelligence – Worldwide | Market Forecast(<)/em(>), Statista, (<)a href='https://www.statista.com/outlook/tmo/artificial-intelligence/worldwide/'(>)https://www.statista.com/outlook/tmo/artificial-intelligence/worldwide/(<)/a(>) (last visited July 14, 2025) [(<)a href='https://web.archive.org/web/20250715163558/https://www.statista.com/outlook/tmo/artificial-intelligence/worldwide'(>)https://web.archive.org/web/20250715163558/https://www.statista.com/outlook/tmo/artificial-intelligence/worldwide(<)/a(>)] (estimating global growth up to $1 trillion by 2031, including consumer facing and enterprise AI products).—is: do AI companies walk the talk when purpose collides with profit? How does their business structure constrain or enable them to do so?

Overall, the report finds that business structure alone is insufficient to ensure companies abide by their mission. Structure does create friction, but capital eventually overcomes it. OpenAI’s transformation speaks to this dynamic. The attorneys general leveraged their jurisdiction over non-profits, citing their ability to block the company’s restructuring to secure commitments to the mission. Be that as it may, the original structure was dismantled precisely because it worked. When the non-profit became an impediment to raising capital, the company conceded some mission safeguards as a compromise to break away from its capital-raising constraints. In fact, financial considerations deeply shape the business structure of all AI companies under review. Through a series of case studies, this research finds that as AI development goes from proof of concept to product, AI companies consistently modify their business structures so as to become more responsive to financial considerations and reduce mission constraints on their actions. This creates deep tension within the market structure for AI: at the moment where these companies are most poised to enact their mission, they evolve their corporate governance to free themselves from any commitment to achieving it.

This suggests that the ability to evolve out of a constraining corporate form is itself a structural defect of our corporate governance system. Private, reversible choices as to the legal framework companies operate under are too brittle to operationalize public missions. Though various stakeholders have mobilized the levers of corporate law to secure some limits on the behavior of AI companies, especially when it comes to attorneys general using their ability to enforce the non-profit form of OpenAI to secure some safety commitments from the company, these efforts remain few and far between. Without binding, capture-proof regulation, corporate mission statements ring hollow. 

This report assesses how various business structures fare in practice in terms of their stated commitments to benefit humanity. Though the contours of the mission of AI companies remain vague, morph over time and vary across companies, at a high level they all signal a commitment to public interest, as opposed to the traditional goal of corporations which consists of maximizing profit for shareholders. Taking that promise as more than newspeak, the case studies below assess whether variations among corporate structures have any bearing on companies’ ability to fulfill their missions. The aim is to surface correlations between how companies are structured and how they behave, comparing how different corporate structures enable them to carry out their stated mission. Confounding variables such as founder culture, investor preferences, and market position stand in the way of a causal relationship between corporate structure and behavior. Rather, the report starts a conversation about corporate structure as one factor shaping how AI labs walk the talk on their mission, and highlights corporate law levers to hold AI labs to their word.

Part I explains the legal business structures of Google DeepMind, OpenAI and Anthropic, outlining their evolution over time. Part II assesses how each company behaves in practice, based on their policy positions on AI regulation, internal safety practices and military contracts. Given the pace of events, the report tracks corporate conduct through February 2026. In fact, the shapeshifting of corporate structures (e.g., OpenAI pivoting from a non-profit to a ‘capped-profit’ and then to a public benefit corporation) is itself an important insight, suggesting a strategic instability that impedes analysis. Nevertheless, these case studies give a sense of how AI companies negotiate tensions between their public mission and self-interested considerations. Part III proposes avenues to demand companies follow through on their mission.

Part I. Taking Stock of Business Structures

AI companies are adept at crafting bespoke legal structures to hardcode public interest in their DNA. Yet as AI went from proof of concept to viable product, they have all diluted such commitments. Over time, AI labs have tended to adapt their corporate form to be more responsive to investors, citing the need to secure massive investments to train and run ever larger frontier models. Google absorbed DeepMind, and, along the way, shed plans for external oversight. Originally a non-profit, OpenAI’s structural changes introduced powerful financial considerations. Anthropic’s structure promotes public interest governance, but it remains brittle as profit-driven shareholders retain the ability to roll back guardrails.

Part II. Business Structures in Action: Assessing Corporate Practices

How do AI companies implement their mission to “benefit humanity”? This section evaluates whether some corporate structures are more conducive to walking the talk on public interest missions by documenting concrete actions in the context of different corporate forms. It finds that overall, there is no meaningful correlation between corporate structure and how companies implement their mission. 

The position of AI companies with regard to proposed regulation gives a sense of their willingness to actually agree to external constraints with compliance costs. While they clamor for regulation in the abstract, they tend to oppose specific regulatory proposals. As for self-governance practices, publicly available information suggests that companies use some self-restraint to put safety above profits, though no independent verification allows a definitive determination on the matter. Finally, initial limitations on AI weapons and military applications gradually eroded across most AI companies, each abandoning earlier restraints. This speaks volumes as to their actual commitment to put principles above profit. 

As Annexes 1–3 show, corporate form is not a stable trait but a moving target, with each company restructuring at its own pace.119Google and OpenAI were corporations and a capped-profit structure throughout the case studies, but Anthropic added the long-term benefit trust after some occurrences. (<)em(>)See(<)/em(>) Annex 1 for detailed timelines. Different market positions and maturity levels further confound a definitive cross-company comparison; the more telling evidence often lies within a single company, where conduct can be tracked against its own structural shifts. Finally, the case studies are necessarily time-bound, with events unfolding almost in real time. Given the dynamism of this space, the sample provides a retrospective assessment of company behavior, not a comprehensive up-to-date analysis. Where relevant, developments subsequent to the February 2026 cutoff date are flagged for future research.

Part III. Levers to Enforce the Mission

Bespoke corporate structures have not always delivered in practice when it comes to prioritizing mission over profit. However, corporate governance offers levers to hold AI companies to their word. This section explores legal tools within the corporate and securities law toolkit, listing the main recourses for each type of corporate structure under review and assessing their potential and limits

Conclusion

AI companies take different corporate forms and evolve over time. This report assessed whether function follows form, in the sense of whether certain business structures lend themselves to carrying out publicly minded missions more so than others. The case studies testing how various corporate vehicles walk the talk on their public missions suggest that corporate form has some bearing on how companies approach regulation, but little impact on safety and military policy. When competitive pressures and lucrative contracts are in play, financial interests prevail irrespective of the legal vehicle under which a company operates. In the competitive market that is AI, corporate structure alone cannot align companies with their mission. In fact, those very structures buckle under investor pressures, as is the case with OpenAI gradually abandoning its non-profit ethos to raise more capital.

When it comes to enforcing publicly oriented missions, recourse is limited. With the exception of preventing OpenAI from completely abandoning its original non-profit mission, there are very few instances of material gains. That is not to say efforts to activate corporate law levers are futile; as even the prospect of trouble can, and likely has, constrained the actions of AI companies. Perhaps counterintuitively, time-tested investor protection for publicly traded stock offers more impactful avenues to force disclosures. Should OpenAI and Anthropic proceed with their plan to go public, they, too, will be subject to increased transparency requirements. Accordingly, one important insight and avenue for future advocacy consists in activating disclosure requirements for publicly listed AI companies.

Three ways forward stem from this report. First, current corporate levers are underutilized. For example, OpenAI PBC’s board duty to exclusively consider the mission pursuant to its deal with the AGs opens the company to more litigation, especially as the company goes public and shareholders can form coalitions to gather the sufficient stock threshold to proceed with a derivative action on behalf of the PBC. That said, this avenue remains speculative as courts remain extremely reluctant to second-guess directors. Consequently, shareholder activism alone will not align companies with their mission. The limits of current legal vehicles call for new structures that actually have a forcing function on corporate behavior regarding mission compliance. PBCs stop short of that, as they allow but do not require companies to put mission above profits. They are meant to enable profit-driven companies to pursue social goals, not the other way around.

Third and perhaps more importantly, accountability requires robust regulation and public infrastructure. We need laws to move the needle towards publicly beneficial AI. Irrespective of their specific form—and without denying differences in degrees between various corporate vehicles—the self-interest of private companies cannot be meaningfully checked without laws that guarantee companies will factor in the costs of non-compliance. Corporate governance alone cannot ensure companies prioritize their mission over profits. Given the collective action problem that arises from competition among them (something key players recognized as early as 2019),346Amanda Askell, Miles Brundage & Gillian Hadfield, (<)em(>)The Role of Cooperation in Responsible AI Development(<)/em(>) (July 10, 2019) (unpublished manuscript), (<)a href='http://arxiv.org/abs/1907.04534'(>)http://arxiv.org/abs/1907.04534(<)/a(>). there is a fundamental mismatch between the incentives of AI companies jockeying for dominance in a capital-intensive market and the broader public interest. As Anthropic’s cofounder Jack Clark put it: “I am pretty skeptical of things that relate to corporate governance because I think the incentives of corporations are horrendously warped, including ours. […] It’s a real weird thing that this is not a government project.”347Dylan Matthews, (<)em(>)The $1 Billion Gamble to Ensure AI Doesn’t Destroy Humanity(<)/em(>), Vox (July 17, 2023), (<)a href='https://www.vox.com/future-perfect/23794855/anthropic-ai-openai-claude-2'(>)https://www.vox.com/future-perfect/23794855/anthropic-ai-openai-claude-2(<)/a(>); (<)em(>)see also(<)/em(>) Deep Ganguli et al., (<)em(>)Predictability and Surprise in Large Generative Models(<)/em(>), (<)em(>)in(<)/em(>) 2022 ACM Conf. on Fairness Accountability & Transparency 1747 (2022), (<)a href='http://arxiv.org/abs/2202.07785'(>)http://arxiv.org/abs/2202.07785(<)/a(>) (noting economic and prestige incentives to move forward even absent adequate assurances). Even OpenAI tried to secure public funding before abandoning its non-profit orientation.348The Ezra Klein Show, (<)em(>)supra(<)/em(>) note 69 (quoting Altman: “A little known fact, we tried to get the public sector to fund us before we went to the capped profit model. There was no interest.”). Policymakers need to step up with AI regulation that motivates companies to align their actions with the public interest, along with substantial investment in civic-minded public AI infrastructure.

Annexes

Annex 1: Google/DeepMind — Structure & Conduct Timeline

Independent startup → absorbed into a for-profit conglomerate

  1. 2010

    Structural Change: Private Limited Company DeepMind founded as an independent startup to “solve intelligence and use it to solve everything else.”

  2. 2014

    Structural Change: Google Acquisition Acquired by Google; sale conditioned on an ethics board to preserve the mission. Public assurances surface against military applications

  3. 2015

    Policy / Governance: Google dissolves the ethics board after a single meeting

  4. Structural Change: Alphabet Reorg Alphabet at the apex of the conglomerate; DeepMind operates as Google subset

  5. 2018

    Case Study – Regulation / Scaling / Military: Military Project Maven backlash: after staff protest, Google publishes AI Principles pledging no weapons

  6. 2020 – 22

    Case Study – Regulation / Scaling / Military: Military Government / defence contracts continue under the AI Principles guardrails

  7. 2020 – 24

    Case Study – Regulation / Scaling / Military: Regulation Urges less restrictive rules in the EU AIA, frames general-purpose regulation as a bad example, opposes external evaluations of high-risk systems

  8. Apr 2023

    Structural Change: Merger Alphabet merges Google Brain and DeepMind into a single division: Google DeepMind

  9. 2024

    Case Study – Regulation / Scaling / Military: Regulation In California, criticizes SB-1074 for pre-training

  10. Jan 2025

    Structural Change: Restructuring Gemini teams moved under single Google DeepMind umbrella; a new ethics council is established, staffed exclusively by Google

  11. Feb 2025

    Case Study – Regulation / Scaling / Military: Military Lifts the ban on weapons and surveillance applications; drops the reference to “harmful applications.”

  12. 2025

    Case Study – Regulation / Scaling / Military: Regulation Supports the federal ban on state AI laws via its trade groups; its global policy president criticizes uneven state-by-state protections

Annex 2: OpenAI — Structure & Conduct Timeline

Non-profit → capped-profit hybrid → public benefit corporation

  1. 2015

    Structural Change: Non-stock Non-profit Incorporated as OpenAI, Inc., a non-profit counterweight to profit-driven labs. Early stance broadly disfavors military applications

  2. 2018

    Policy / Governance: Non-binding Charter commits to a “fiduciary duty to humanity.”

  3. 2019

    Structural Change: Capped Profit + LLCs Pivot to a “capped-profit” structure to raise capital; Microsoft partnership begins

  4. 2019 – 24

    Case Study – Regulation / Scaling / Military: Military Prohibits military and warfare applications and weapons development

  5. 2023

    Policy / Governance: Microsoft invests $10B+ for 49% of profits; non-profit keeps formal control. Board fires Altman, then reinstates him within days

  6. 2021 – 23

    Case Study – Regulation / Scaling / Military: Regulation Grow increasingly critical of the EU AIA; threatens to withdraw ChatGPT from EU

  7. 2024

    Case Study – Regulation / Scaling / Military: Military Drops the blanket “military” ban, pursues defence contracts

  8. 2024

    Case Study – Regulation / Scaling / Military: Regulation Opposes SB-1047 and state regulation

  9. 2024

    Policy / Governance: Tender offer at a $157B valuation lets employees sell profit-participation units

  10. 2025

    Case Study – Regulation / Scaling / Military: Regulation Pushes for federal preemption; and advocates for a non-binding federal framework

  11. 2023 – 26

    Case Study – Regulation / Scaling / Military: Scaling / Safety Preparedness framework commits to pre-deployment safeguards

  12. Oct 2025

    Structural Change: Public Benefit Corp Restructured as a Delaware PBC; non-profit retains control and owns 26% of stock; valuation ~$300B

  13. Feb 2026

    Case Study – Regulation / Scaling / Military: Military Offers product to Pentagon for all lawful uses hours after Anthropic’s refusal

Annex 3: Anthropic — Structure & Conduct Timeline

PBC with a Long-Term Benefit Trust

  1. Jan 2021

    Structural Change: Delaware PBC Founded by former OpenAI staff (incl. Dario Amodei) over safety concerns; incorporated as a Delaware public benefit corporation from the outset

  2. 2021 – 2023

    Case Study – Regulation / Scaling / Military: Regulation Public commentary supportive of risk-based regulation; engages EU on high-level safety goals

  3. 2023

    Structural Change: + Long-Term Benefit Trust Disinterested trustees take a long view on safety, but profit-driven shareholders retain the ability to roll back the trust

  4. Sep 2023

    Case Study – Regulation / Scaling / Military: Military First archived Acceptable Use Policy bans weapons development and “battlefield management applications”, but never military use as such

  5. Jun 2024

    Case Study – Regulation / Scaling / Military: Regulation Files a “support-if-amended” letter on SB-1047: opposes prescriptive pre-harm mandates, proposes outcome-based deterrence

  6. Case Study – Regulation / Scaling / Military: Military Usage Policy adds protection of military bases; exceptions allow carve-outs for government customers

  7. Aug 2024

    Case Study – Regulation / Scaling / Military: Regulation Cautious support SB-1047 after amendments, as benefits likely outweigh its costs

  8. Nov 2024

    Case Study – Regulation / Scaling / Military: Military Partners with Palantir and AWS to bring Claude to US intelligence and defense agencies in classified environments

  9. May 2025

    Case Study – Regulation / Scaling / Military: Scaling / Safety Responsible Scaling Policy v2.2 adopts precautionary approach; a high-water mark for voluntary safety measures

  10. Jun 2025

    Case Study – Regulation / Scaling / Military: Regulation Opposes the proposed 10-year federal moratorium on state AI regulation as “too blunt an instrument”, proposing a federal transparency standard instead

  11. Feb 2026

    Case Study – Regulation / Scaling / Military: Scaling / Safety Walks back safety commitments, citing competitive pressure

  12. Case Study – Regulation / Scaling / Military: Military Conflict with Pentagon over autonomous weapons and mass surveillance

Credits

Thanks to Aashna Agarwal, Sacha Alanoca, Kate Brennan, Jesse Fried, Shira Gur-Arieh, Teodora Groza, Heidy Khlaaf, Martha Minow, Sarah Myers West, Robert Yalden and Tom Zick for comments and feedback, along with Camille Chichet, Oliver Kulik, Nalanda Sharadjaya and Mariana Ramirez Cabal for research and editorial assistance. 

The author gratefully acknowledges financial support from the Harvard Law School Summer Public Interest Funding program.

Design by Partner & Partners.